Research Analyze

Confidence levels:

• High: 3+ strong independent sources, no significant contradictions
• Medium: 2 sources, or mixed quality, or minor contradictions
• Low: Suggestive but evidence is thin — flag this clearly

Single-source observations stay in the source note, not in themes.md.

Building contradictions.md

A contradiction belongs here when two sources make meaningfully different claims about the same thing in ways that matter to the research question.

For each contradiction:

## Contradiction N: [Topic of disagreement]

**Source A**: S00N claims [X]
**Source B**: S00N claims [Y]

**Possible explanation**: [Different time periods? Different geographies?
Different definitions? Different methodologies?]

**Resolution**:
- [ ] Resolved — [explanation]
- [ ] Unresolved — will be disclosed as limitation

**Impact on research**: high / medium / low

Rules:

• Never resolve a contradiction by silently picking one source
• If unresolved, it must surface in the output as a limitation
• A contradiction that can be explained (e.g. different time periods) is resolved

Updating gaps.md

After reading all sources, update the gaps file:

• Close gaps that sourcing has filled — move to "Closed gaps" section
• Add new gaps discovered during analysis — things sources implied but didn't answer
• Reclassify any sourcing type gaps that are actually inherent limitations (i.e. the question may not be answerable with available evidence)

Gap entry format:

### G00N: [Description]
**Type**: sourcing / clarity / inherent / out-of-scope
**Priority**: high / medium / low
**Related question**: Q[N]
**What we have**: [what sources say so far]
**What's missing**: [what would actually answer the question]
**Status**: open / closed

Deciding whether to stop or get more sources

If you find a critical gap — a main research question from plan.md has zero sources addressing it — stop and tell the user:

"Gap G00N is critical and unaddressed. Recommend using /research-source focused on [specific search strategy] before completing analysis."

If gaps are peripheral or the research question has adequate coverage → continue.

Finishing analysis

When all active sources are read and all three analysis files updated:

Write a summary paragraph at the top of themes.md:

"This analysis identified N themes across N active sources. The strongest finding is [X]. The most significant unresolved contradiction is [Y]."

Update state.json:

• phase: synthesis
• progress.analysis_complete: true
• last_action: "Analysis complete — N themes, N contradictions, N gaps identified"
• next_action: "Begin synthesis — use /research-synthesize"
• Add history row